Simple Improved Reference Subtraction for H4RG, H2RG, and H1RG Near-infrared Array Detectors

TL;DR

Simple Improved Reference Subtraction (SIRS) presents a least-squares, reference-pixel-based method to suppress correlated read noise in Teledyne HxRG detectors, applicable to H4RG, H2RG, and H1RG arrays. It extends IRS^2 ideas from JWST to conventional clocking patterns, using incomplete real Fourier transforms and training data to produce an optimal, linear correction without tuning. Validation on Roman ground-test data demonstrates substantial reductions in low-frequency banding and correlated noise while preserving mean detector response; the method is implemented in Julia with a Python backend for practical use, including archival data applicability. This work provides a practical, statistically optimal framework for reference-based noise subtraction that can adapt to evolving detector hardware and observing modes.

Abstract

Teledyne's H4RG, H2RG, and H1RG near-infrared array detectors provide reference pixels embedded in their data streams. Although they do not respond to light, the reference pixels electronically mimic normal pixels and track correlated read noise. In this paper, we describe how the reference pixels can be used with linear algebra and training data to optimally reduce correlated read noise. Simple Improved Reference Subtraction (SIRS) works with common detector clocking patterns and, when applicable, relies only on post-processing existing data so long as the reference pixels are available. The resulting reference correction is optimal, in a least squares sense, when the embedded reference pixels are the only references and the reference columns on the left and right are treated as two reference streams. We demonstrate SIRS using H4RG ground test data from the Nancy Grace Roman Space Telescope Project. The Julia language SIRS software is freely available for download from the NASA GitHub. The package includes a python-3 ``backend'' that can be used to apply SIRS corrections if a SIRS calibration file has been provided by the instrument builders.

Figures (9)

• Figure 1: We developed SIRS for use with Roman ground test data. a) Roman's Wide Field Instrument (WFI) will fly eighteen $4096\times 4096$ pixel Teledyne H4RGs. b) The WFI configuration uses 32 outputs as shown here. All HxRGs embed a 4-pixel wide border of reference pixels in the data streams. The Roman fast scan directions use the default Teledyne configuration that alternates. In this figure, the fast scanners run from left to right in the odd numbered white colored outputs and from right to left in the even numbered grey colored outputs. The pixel readout rate is $2\times10^5~\textrm{pixels}~\textrm{s}^{-1}~\textrm{output}^{-1}$. There is a few pixel time overhead at the end of each row before clocking the first pixel of the next row. The prototype H4RG-10 shown in panel a is included for illustrative purposes only. It is not the flight design and some details have been deleted to comply with U.S. legal requirements. This figure should not be relied upon for engineering purposes.
• Figure 2: Panel a) shows the power spectrum on a log-log plot, in instrumental units, for two different outputs. We saw no qualitative differences between any outputs. Instrumental units are sufficient for making relative comparisons. SIRS is highly effective in rejecting the $1/f$ noise that appears at low frequency. Other notable features include bands at 60 Hz and harmonics due to the U.S. power grid and b) a "comb" pattern spaced at intervals of the line frequency arising from interpolation over the regularly spaced end-of-line gaps. Panel b) shows the same points on a semilog scale. In addition to the features already mentioned, there is a line (of undetermined origin) at 83,688 Hz and its out of band second harmonic appears aliased down at 32,624 Hz. We looked for its third harmonic aliased down to 51,064 Hz, but did not find it. This suggests that either it is not very strong, or it may have been blocked by filters in the readout electronics. We tentatively attribute the enhanced "comb" pattern near 100 kHz to out of band "comb" aliased down.
• Figure 3: Panel a) shows the covariance of the row-medians of output #16 computed in the time domain. There is strong off-diagonal covariance, indicating that the different time steps (pixel indices) are correlated in this representation. The diagonal emerges much more clearly in b) Fourier space. Both panels a and b are on the same greyscale. The clouds around the edges in panel b reflect at least in part the uncertainties in the measurement. They become less pronounced as more data is added. The clouds are caused by the $1/f$ noise that SIRS aims to remove. The feature at 120 Hz requires more study, but it may appear so clearly because this line is so much stronger than any others (see Figure \ref{['fig:npsd']}).
• Figure 4: This figure shows $\bm{\mathbf{\alpha}}$ and $\bm{\mathbf{\beta}}$ for 3 of H4RG 20663's 32 outputs. Output #1 is the left-most output, output #16 is just left of center, and output #32 is the right-most output. As expected, the relative weights taken by $\bm{\mathbf{\alpha}}$ and $\bm{\mathbf{\beta}}$ shift depending on which reference columns are closest. Broadly speaking, outputs 2-15 look intermediate between outputs 1 & 16 and similarly outputs 18-31 look intermediate between outputs 17 & 32.
• Figure 5: a) This figure re-plots Fig \ref{['fig:sirs-alpha-beta']} (center) on a linear scale. Visible features include $1/f$ noise and lines and bands associated with the US power grid at 60, 120, 180, and 240 Hz. c) Because the reference pixels are the first pixels read in every line, the absolute value of the phase increases from zero at 0 Hz to $\frac{\pi}{2}$ at $f_\textrm{Ny,line}$. b) There is very little amplitude near the Nyquist frequency, $f_\textrm{Ny}=200~\textrm{kHz}$, and d) phase is almost entirely noise at these frequencies. We therefore zero-out frequencies $f>f_\textrm{Ny,line}$ when working with the existing Roman data.